Gates

Gates apply intensity thresholds to classify cells as marker-positive or marker-negative:

1. Select measurement — Choose the intensity measurement to gate on (e.g., mean CD3 intensity in the cytoplasmic compartment).
2. Set threshold — Place the gate at the intensity value separating the positive and negative populations. This can be done manually (examining the histogram) or automatically.
3. Classify — Every cell with intensity above the gate is marked positive; every cell below is marked negative. The result is a binary column in the measurement table.
4. Validate — Overlay the gate result on the image to visually verify that positive cells correspond to truly stained cells and negative cells correspond to unstained cells.

The Rose criterion applied to gating: Pawley's Rose criterion states that a signal must exceed 5× the noise to be reliably detected. Applied to gating: a truly positive cell must have intensity exceeding the negative population mean + 5× the negative population's standard deviation to be reliably classified as positive. Cells in the 2-5σ range above the negative mean are in the "equivocal zone" — they might be dim positives or bright negatives.

Statistical thresholding theory (Dilbilir): Setting a gate is a special case of binary classification where the feature is one-dimensional (intensity) and the decision boundary is a single threshold. The optimal threshold minimizes total classification error — the sum of false positives (negative cells above the gate) and false negatives (positive cells below the gate). When the positive and negative populations have different sizes, the optimal threshold shifts: in a sample where 90% of cells are negative and 10% are positive, a slightly lower threshold (catching more positives at the cost of some false positives) may minimize total error.

The Gaussian intersection: If the positive and negative populations have Gaussian intensity distributions with means μ₊ and μ₋ and standard deviations σ₊ and σ₋, the optimal threshold falls at or near the intersection of the two Gaussians. When σ₊ = σ₋, this is the midpoint (μ₊ + μ₋)/2. When variances differ, the intersection shifts toward the population with larger variance. This is the Bayesian optimal threshold for equal misclassification costs.

Batch effects on gates: Staining intensity varies between batches — even with the same protocol, day-to-day variation in stain concentration, incubation time, and washing steps shifts the positive and negative populations. Gates set on one batch may not apply correctly to another. Two approaches: (1) set gates per-sample by examining each sample's histogram, or (2) normalize intensity across batches before gating. Per-sample gating is more accurate but labor-intensive; normalization enables consistent gates but assumes the batch effect is the same for all cell types.

Setting gates for a multiplex IF panel:

1. CD3 gate: Bimodal histogram with clear valley at intensity 45 → set gate at 45
2. CD8 gate: Partially overlapping populations → set gate at 60 (valley center) → flag 5% of cells as equivocal (within ±10 of gate)
3. PD-L1 gate: Continuous distribution without clear valley → set gate at negative mean + 3σ → binary classification less reliable, report as "PD-L1 scoring" rather than strict positive/negative

The reliability of each gate depends on how well the positive and negative populations separate. CD3 gates are usually clean; PD-L1 gates are often challenging because expression is a continuum rather than a binary on/off.